Re: [v6ops] PMTUD issue discussion

[Fernando Gont <[e-mail]>]

On 08/26/2014 12:42 PM, Templin, Fred L wrote:
>> On 08/26/2014 12:13 PM, Fred Baker (fred) wrote:
>>>> I think it would be useful to see the discussion of this measure
>>>> and its applicability/tradeoffs in the draft.
>>>
>>> It seems like it might have value to also test the interaction
>>> with various initial window settings, and look at its inclusion in
>>> other relevant OS’s including FreeBSD, Windows, and MacOSX.
>>
>> FWIW, while RFC4821 can work without relying on PLPMTUD at all, some
>> OSes (such as Windows) only fall back to ICMP-less PMTUD as part of
>> "ICMP blackhole detecion" -- and they have been doing this for years
>> now...
> 
> I think the point is that PLPMTUD is not well supported in modern
> OSes in the spirit in which it was specified in RFC4821. That
> needs to change.

FWIW, I'm all in favor of RFC4821, since it improves robustness of PMTUD.

That said, as far as I can recall, using RFC4821 for blackhole detection
(when traditional PMTUD fails) is well within the spirit of RFC4821
(RFC4821 doesn't push ICMP-less as the only implementation strategy).

What I recall from discussing this with Linux developers circa 2005 is
that a complete ICMP-less PMTUD is rather undesirable, since it tends to
have longer convergence time when compared with "traditional PMTUD" ||
"Traditional PMTUD + RFC4821 for blachole detection".

Thanks,
-- 
Fernando Gont
SI6 Networks
e-mail: [e-mail]
PGP Fingerprint: 6666 31C6 D484 63B2 8FB1 E3C4 AE25 0D55 1D4E 7492